Sensitive Continuous Monitoring of thanks to Matrix of several Suspended Gate Field Effect Transistors B. da Silva Rodrigues a,b, O. De Sagazan a, S. Crand a, F. LeBihan a, O. Bonnaud a, T. Mohammed-Brahim a, N.I. Morimoto b a GM-IETR UMR-CNRS 6164, Universite RENNES I, 35042 Rennes Cedex, FRANCE b LSI, Universidad de São Paulo brahim@univ-rennes1.fr Introduction Water quality becomes now a major world problem so that there is a high demand in sensitive -meter allowing continuous monitoring of the global quality of water. Indeed is a good indicator of this global quality. For example plant proliferation in water leads to increase. Usual glass -meter is more and more replaced now by Ion Sensitive Field Effect Transistors (ISFETs). Since Bergveld [1] first employed the field effect transistor in neurophysiological measurements in 1970, ISFETs have been developed into a new type of chemical sensing electrode. Many theoretical and experimental studies have been published for describing the behavior of this chemical-sensing electronic device [2,3]. As transistors, ISFETs are fabricated by the usual microelectronic technology leading to miniaturization, cost lowering and integration with electronics for signal treatment. The structure of ISFETs is based on the classical MOSFET (metal oxide field effect transistor). Only the usual metallic gate contact is removed. Aqueous solution with a reference electrode dipped in, will serve as gate of this particular transistor. The sensitivity of ISFET is explained from the shift of its threshold voltage induced by the variation of the flat-band voltage, V FB, when the distribution of the electrical charges inside the solution varies. Then, any variation of the H + ion concentration that governs the of the solution, will lead to threshold voltage shift. The shift is explained by the linear variation of the surface potential at the insulator-solution interface ψ 0 with the of the solution [4] : ψ 0 / = -2.3KT/qα where T is the temperature, K the Boltzmann constant, q the electron charge and α a dimensionless parameter, varying between 0 and 1. In this case, the maximum sensitivity, which is the Nernstian sensitivity, is -59 mv/ at 20 C, when α equal 1. So, the sensitivity of all the present ISFETs is lower than, or close, to this maximum. In many cases, higher sensitivity is needed. For example, drinking water monitoring needs the measurement of less than 1 variation. In other fields, cell culture monitoring in biology or chicken farming monitoring need high sensitive measurements. Sensitivity can be increased by using suspended gate at submicronic distance from the channel of the transistor [5]. This submicronic distance allows the field effect to increase the sensitivity more than becomes pore than 4 times the Nernstian sensitivity [6]. The
new structure was called Suspended Gate Field Effect Transistor (SGFET). It was originally developed by Janata [7]. Other variations of the structure followed (Hybrid SGFET [8], Floating Gate FET [9]). These first structures were developed only in the goal to fix the usual reference electrode of ISFETs. Indeed, the suspended gate was too far from the transistor channel to profit from the field effect. Using new microtechnology tools, we were able to decrease the gap still submicronic values and to increase highly the sensitivity. Here, technological improvements of the structure are presented in the goal to reach practical, simple to use, highly sensitive -meter, able to monitor the of the water and its quality consequently. SGFET process and sensor fabrication A scheme of P-type SGFET just before removing the germanium sacrificial layer is presented in Figure 1<100> oriented and low-doped N type silicon wafers are used as substrate. The drain and source wells come from nitride boron diffusion at 1100 C with a 650 nm SiO 2 as protect layer. The channel is then oxidized at 1100 C so that the SiO 2 thickness is 70nm. After that, the channel oxide is protected by a 50nm Si 3 N 4 LPCVD (Low Pressure Chemical Vapor Deposition) layer deposited at 600 C. On the silicon nitride layer, a 500nm USG (Undoped Silicat Glass) sacrificial layer is deposited by PECVD at 350 C, followed by a 50 nm silicon nitride layer, which acts as isolation layer for the bottom of the suspended gate. Al Poly Si P+ SiO 2 Ge USG Si 3 N 4 Si P+ BulkSi N Figure 1: Scheme of the structure of P type SGFET before the releasing of the USG sacrificial layer. Figure 2: SEM micrograph of the SGFET structure Gate and drain source contacts are then obtained by the deposition of 500nm thick P-type polysilicon layer amorphously deposited at 550 C and crystallized during 12h at 600 C. A last 50 nm nitride layer protects the top of the gate. Vias are opened in this silicon nitride layer to realize contact with Al path. Due to the liquid ambient use, all metal layers are protected with oxide layers such BPSG (Boro Phospho Slicat Glass) and nitride PECVD layers. The last step of the process is the releasing of the gate by a HF etching of the USG sacrificial layer. Scanning Electron Microscope view of the final structure is shown in Figure 2.
The total area of the SGFET transistor, including the connections is 130µm x 70 µm. This small area led to integrate locally many similar SGFETs in a square matrix configuration. SEM micrograph of such matrix is shown in Figure 3. The strong integration on small area allows the SGFETs of the matrix to be dipped in the same medium and then to present the same characteristic in ideal case. Experimental characteristics of all SGFETs can be measured and averaged. By this means, redundancy is introduced in the goal to reach the maximum reliability. Figure 3: SEM micrograph of matrix of SGFETs showing some transistors Figure 4: Bonded on PCB card SGFET matrix as ready to use sensor. For practical use, SGFET matrix is bonded in a PCB card (Figure 4). All wires are coated in special resin to avoid short circuits due to the liquid immersion of the device. The bonded matrix can be considered as ready to use sensor. The electrical characterization of all the transistors of the matrix is performed with an Agilent B1500. During the measurements, the PCB card is dived into the solution under analyzing. measurements Different experiments have been done to test the ability of the matrix sensor in monitoring. The first one was the sensor answer when the value is suddenly varied by a large amount. The of the solution was initially 3 and the sensor is dipped in. The transistors of the matrix-sensor were polarized with a gate voltage of -3V and a drainsource voltage of -0.05V. The drain-source current was monitored continuously. After some time, a solution at =12 was added suddenly to the first one. Figure 5 presents the time behaviour of the drain-source current before and after the solution mixing. The current is more or less stable before the mixing, then increases suddenly and stabilized finally. It is not easy to determine the response time from this experiment as the adding of the second solution was made manually and then the mixing time was not controlled. However, its evaluation gives an order of 200 sec.
-8-12 -16-20 -24 20 40 60 80 100 time (minutes) 10 8 6 4 2 120 100 80 60 40 20 0 20 40 tim e (m inutes) 8 6 4 2 0 Figure 5: Drain-source current monitoring at constant value (3), then after the sudden adding of a solution of a of 12 (full plot). The values (open plot) were measured by a macroscopic commercial -meter. Figure 6: Drain-source current monitoring at increased still 40 minutes and then decreased (full plot). The values (open plot) were measured by a macroscopic commercial -meter. This first experiment showed the possibility to monitor variation. This monitoring was well highlighted in the second experiment where the of acid solution was continuously and manually increased by adding increased content of basic solution still 40 minutes and then decreased by added increased content of acid solution. During this experiment, the transistors of the matrix-sensor were polarized with a gate voltage of -3V and a drain-source voltage of -0.3V. SGFETs are able to monitor increasing or decreasing of solutions. 140 120 100 80 60 40 20 3 4 5 6 7 8 Figure 7: Linear increase of the drain-source current with. This monitoring can be characterized by the sensitivity of the current to values. From the data of this second experiment, the current can be plotted versus values (Figure 7). In spite of the dispersion of the experimental points that is due to the manual, and then
not well controlled, variation, linear regression gives a sensitivity of 13 µa/. This sensitivity means that a variation of 0.1 can be electronically detected as a variation of 1.3 µa is easily measurable. It is not easy to compare the present current sensitivity to the performance of usual meters and those found in the literature. Indeed, sensitivity of glass or ISFET -meters is usually expressed in terms of voltage variation as the surface potential varies with in both techniques. Comparison can be done if the transfer characteristics, drain-source current versus gate voltage, of SGFETs are measured at different values. So, of acid solution was varied by adding small content of basic solution and after waiting some seconds, transfer characteristics of SGFETs were measured. The experiment was renewed at increasing values. Figure 8 shows these transfer characteristics at different values. If we do not consider the abnormal behaviour of the curves at low current values that is due to parasitic transistor, the curves shift towards positive gate voltage when increases. When SGFETs are really in their on-regime, the shift of the gate voltage at a drain current of 25 µa can be measured. It is plotted in Figure 9. The linear shift led us to determine a sensitivity of 400 mv/, much higher than its value for the usual meters. Indeed the sensitivity of usual -meters is limited by the theoretical Nernstian value that is 59 mv/ at 20 C. increase 30 20 10 0 2 1 0-1 -2-3 Gate Voltage (V) Shift of V G at I D =-25µA (V) -1.2-0.8-0.4 0 8 7 6 5 Figure 8: Transfer characteristics of P-type SGFET at different increasing values. Figure 9: Shift of the gate voltage V G at fixed drain current to -25µA when value of the solution, where SGFETs are characterized, varies. Optimisation of the electric measurements Through these different electrical tests it has been noticed that measure repeatability in the sampling mode could soon became a limit of the system. Different ways of testing have been investigated to increase the reliability of measures. One of our hypotheses was that during a long sampling measure the characteristic of the devices drift slowly due to the stress of the measure. That is the reason why the figure.7 presents some points quite far from the linear regression of the sensitivity. To prevent the drift due to the sampling measure, time of the measure has been reduced to 5 seconds (figure. 10) and the final drain current value is the averaged Ids current on this period. The of different solution
has been measured by this method. For each seven sampling of 5 second have been performed. The figure 11 presents the device sensibility calculated by this way. The value of 10.4µA/ is close from the former one calculated on figure.7 and the statistic dispersion of measures has been widely decreased. -0,00019-0,00018-0,00017-0,00016-0,00015-0,00014-0,00013 3.95 4.85 5.65-0,00012 0 1 2 3 4 5 Time (s) Figure 10: Drain current sampling of 5 seconds for different value of the. The final result of Ids is given by the average of the Ids on the 5 seconds. -180-175 -170-165 -160-155 -150 3,8 4,0 4,2 4,4 4,6 4,8 5,0 5,2 5,4 5,6 5,8 Figure 11: Sensibility of the drain-source current with. At each seven measures have been performed allowing to estimate the statistic reliability of the method and the device. Sensibility obtained by this way is 10.4µA/.
Conclusion An SGFET-type charge sensor matrix has been processed by standard microelectronics technology and bonded on a PCB card in order to realise a full electronic -meter. The detection principle is based on the field effect mechanism of MOSFET. Detections have been performed thanks two different electrical methods. Both have given interesting results. Sampling methods have proved its reversibility and its effectiveness for monitoring. Sensitivity of 13µA/ have been found, which could fit with several applications especially the water quality monitoring. The sweep measurement method which is more complex to adjust has given sensitivity of 400mV/ in the same domain of, this result shows particularly the high level of sensitivity induced by the SGFET technologies. Electric measures reliability have been improved by optimising the way of test, limiting the sampling time and averaging the drain current value on this small period. Next experiments will try to increase the reliability of the measure by development of specific cleaning and rinse protocols. Furthermore we are going to develop a microcontroller device to monitor more sharply the measure protocol with the SGFET matrix. In fact one of the main advantages of an electronic -meter is the possibility to integrate near the sensor in the silicon chip electronic functions for the signal treatment and the data storage. Acknowledgments Authors would like to thanks Christian Le Mouellic, Patrice Lancelot Dominique Bocquene and Joseph Tregret from MHS (Nantes-France) for their help and their manufacturing advices in the project and CAPES and CNPq for the financial support. References [1] P.Bergveld, Development of an ion sensitive solid-state device for neurophysiological measurements, IEEE Trans. Biomed. Eng.17 (1970) 70-71. [2] Clifford D. Fung, Peter W. cheung, Wen H. Ko, A generalized theory of an electrolyte insulator semiconductor field effect transistor, IEEE Trans. Electron Devices 33 (1) (1986) 8-18. [3] P. Bergveld, Thirty years of ISFETOLOGY What happened in the past 30 years and may happen in the next 30 years, Sensors and Actuators B, 88 (2003) 1-20. [4] R.E.G. van Hal, J.C.T. Eijkel, P. Bergveld, Adv. Colloid. Interface Sci. 69, 31 (1996). [5] F. Bendriaa, F. Le Bihan, A.C.Salaün, T. Mohammed-Brahim, O. Bonnaud, SPIE s International Symposium on Microtechnologies for the New Millennium, Sevilla, Spain, (2005) [6] F. Bendriaa, F. Le Bihan, A.C.Salaün, T. Mohammed-Brahim, O. Bonnaud; Int. Conf. Sensing Technology ICST, Palmerston North, New Zealand, November 21-23, 2005 [7] J. Janata; Sensors and Actuators, 4 (1983), pp. 255-265. [8] B. Flietner, T. Doll, J. Lechner, M. Leu, I. Eisele, Sensors and Actuators B 18-19 (1994) pp. 632-36. [9] M. Burgmair, H.P. Frerichs, M. Zimmer, M. Lehmann, I. Eisele, Sensors and Actuators B 95 (2003) pp. 183-188.